KR100195123B1 - Image Quality Improvement Method Using Average-Matched Histogram Equalization of Low-pass Filtered Signal and Its Circuit - Google Patents

Abstract

The method of improving image quality using average-matched histogram equalization of low-pass filtered signal outputs low-pass filtered signal by low-pass filtering the input video signal, and histogram equalization using low-pass filtered signal cumulative density function as a transform function. Adjust the conversion function so that the average level of the obtained signal is mapped to itself, output the improved signal, obtain the difference from the low-pass filtered signal from the input video signal, and add the obtained difference value to the improved signal to output the output signal. As a result, noise is improved when the contrast of a given video signal is improved, thereby improving image quality.

Description

Image Quality Improvement Method Using Average-Matched Histogram Equalization of Low-pass Filtered Signal and Its Circuit

The present invention relates to a method for improving image quality using a mean-matching histogram equalization of a low-pass filtered signal and a circuit thereof, and more particularly, to low-pass filtering an input image and to improve contrast by using a mean-matching histogram equalization of a low-pass filtered signal. At the same time, the present invention relates to a method and a circuit for improving image quality by suppressing noise from improving in contrast.

The basic operation of histogram equalization is to convert a given input image based on the histogram of the input image, where a histogram represents a gray level distribution in a given input image.

This gray level histogram provides an overall depiction of the appearance of the image. Gray levels appropriately adjusted according to the sample distribution of the image improve the appearance or contrast of the image.

Among many methods for improving contrast, histogram equalization, the method of improving the contrast of a given image according to the sample distribution of the image, is the most widely known and described in [1] and [2]: [1] JSLim , Two-Dimensional Signal and Image Processing, Prentice Hall, Englewood Cliffs, New Jersey, 1990, [2] RC Gonzalez and P. Wints, Digital Image Processing, Addison-Wesley, Reading, Massachusetts, 1977.

In addition, useful applications of histogram equalization methods, including medical image processing and radar image processing, are disclosed in [3] and [4] below: [3] J. Zimmerman, S. Pizer, E. Staab, E. Perry, W. McCartney, and B. Brenton, Evaluation of the effectiveness of adaptive histogram equalization for contrast enhancement, IEEE Tr.on Medical Imaging, pp. 304-312, Dec. 1988, [4] Y.Li, W. Wang , and DYYu, Application of adaptive histogram equalization to x-ray chest image, Proc. of the SPIE, pp. 513-514, vol. 2321,1994.

Therefore, the technique using a histogram of a given image has been usefully applied in various fields such as medical image processing, infrared image processing, and radar image processing.

In general, because histogram equalization has the effect of stretching the dynamic range, histogram equalization flattens the gray distribution of the resulting image, and as a result improves the contrast of the image.

This characteristic of the well-known histogram equalization may be a drawback in practical cases. That is, since the background noise on the video signal is histogram equalized due to the characteristic of improving the contrast by increasing the dynamic range, the background noise is also improved, and thus there is a problem that the image quality is deteriorated especially in a region having a constant level.

An object of the present invention is to perform a low pass filtering on an input image including Gaussian type noise, and when the histogram is equalized using the cumulative density function of the low pass filtered input image as a transform function, the average level of the low pass filtered input image is mapped to itself. The present invention provides a method of improving image quality by adjusting a conversion function to improve contrast and suppressing noise.

Another object of the present invention is to perform a low pass filtering of an input image including Gaussian type noise, and when the histogram equalization is performed using the cumulative density function of the low pass filtered input image as a transform function, the average level of the low pass filtered input image is applied to itself. The present invention provides an image quality improvement circuit that adjusts a conversion function to be mapped to improve contrast and to suppress noise from being improved.

In order to achieve the above object, there is provided a method for improving image quality by using histogram equalization of a video signal represented by a predetermined number of gray levels: (a) by low pass filtering an input video signal. Outputting a low pass filtered signal; (b) outputting an improved signal by performing a histogram equalization using the cumulative density function of the low pass filtered signal as a conversion function and adjusting the conversion function such that the average level of the low pass filtered signal is mapped to itself; obtaining a difference from the low pass filtered signal from the input image signal; And (d) adding the obtained difference value to the improved signal to output an output signal.

In order to achieve the above another object, the image quality improvement circuit according to the present invention is a circuit for improving image quality by using histogram equalization of a video signal represented by a predetermined number of gray levels: Low pass filtering means for outputting a filtered signal; A mean-matching histogram equalization means for outputting an improved signal by adjusting the transform function such that the cumulative density function of the low-pass filtered signal is used as a transform function and the average level of the low-pass filtered signal is mapped to itself. ; Detection means for detecting a difference between the low-pass filtered signal in the input video signal; And adding means for adding the detected difference value and the improved signal to output an output signal.

1 is a block diagram of an image quality improvement circuit using mean-matching histogram equalization of a low pass filtered signal according to the present invention.

FIG. 2 is a detailed circuit diagram of an embodiment of the image quality improvement circuit shown in FIG. 1.

First, the image quality improvement method proposed by the present invention will be described.

The input image composed of _L gray levels {X ₀ , X ₁ , ..., X _L-1 } is called {Y (i, j)} and the low-pass filtered image of the input image {Y (i, j)} Let the input image be {A (i, j)}. Then, a signal having improved contrast (hereinafter referred to as an improved signal) using mean-matching histogram equalization is called A _eq (i, j), and a final output signal is Y _H (i, j).

Accordingly, the final output signal Y _H (i, j) of the present invention is a signal obtained by subtracting the low-pass filtered signal A (i, j) from the input signal Y (i, j). It is the result of adding to A _eq (i, j)), which is given by Equation 1 below.

[Equation 1]

Here, Y (i, j)-A (i, j) correspond to the high frequency components of the original signal.

The basic concept and usefulness of the present invention can be easily understood from Equation 1 representing the output signal Y _H (i, j) as follows.

[Equation 2]

Equation 2 shows that the difference between the final output signal Y _H (i, j) and the improved signal A _eq (i, j) is equal to the input signal Y (i, j) and the histogram pre-equalization signal, that is, the low range. It is equal to the difference from the filtered signal.

This means that if two different samples have the same low-pass filtered value, the difference between these two samples will be constant at the input or output of the present invention. In other words, when two adjacent samples in the image region having a constant signal level have different values due to Gaussian noise, assuming that the low-pass filtered values of these two samples are the same, the difference between these two samples is different from that of the present invention. It is shown that it is kept constant at the input stage and the output stage, which means that the present invention suppresses the noise improvement when the contrast improvement using the mean-matching histogram equalization is improved.

It is well known that low-pass filtering can extract a constant signal level when Gaussian noise is mixed in a video signal having a constant signal level.

To further explain the above description, assume that input sample Y (i, j) and input sample Y (i ', j') have the same low-pass filtered value. That is, suppose A (i, j) = A (i ', j'). Therefore, A _eq (i, j) = A _eq (i ', j').

The output signals corresponding to Y (i, j) and Y (i ', j') are given as follows.

[Equation 3]

[Equation 4]

If A (i, j) = A (i ', j') and A _eq (i, j) = A _eq (i ', j')

[Equation 5]

Can be obtained.

Equation 5 above means that the difference between the two input samples is constant at the input and output of the present invention, which indicates that the present invention suppresses the noise improvement when the contrast is improved by using the average-matched histogram equalization.

Next, the mean-matching histogram equalization algorithm used in the present invention will be described.

The low pass filtered input image {A (i, j)} consists of L discrete gray levels {X ₀ , X ₁ , ..., X _L-1 }, where X ₀ = 0 is black Level, X _L-1 = 1 indicates white level. In addition, X _m represents the average level of the low pass filtered input image {A (i, j)}, and X _m ∈ {X ₀ , X ₁ ,..., X _L-1 }.

The probability density function (PDF) is defined as follows.

[Equation 6]

, for k = 0,1, ..., L-1

Here, n _k represents the number of times the gray level (X _k ) appears in the low-pass filtered input image {A (i, j)}, and n is the total number of samples of the low-pass filtered input image {A (i, j)}. Indicates. At that time, the cumulative density function (CDF) is defined as follows.

[Equation 7]

Based on the cumulative density function, the output after typical histogram equalization (Y _t ) is given as follows for the low-pass filtered signal A (i, j) = X _k .

[Equation 8]

The biggest problem with this typical histogram equalization is that the average brightness between input and output signals can vary significantly depending on the cumulative density function used as the conversion function.

In order to solve this problem, the present invention proposes the following mapping operation based on the average of the low-pass filtered input image in combination with the cumulative density function of the output A _eq (i, j) after the mean-matching histogram equalization.

[Equation 9]

This means that the input sample below the mean (X _m ) Input samples that are mapped to gray levels from X ₀ to X _m by more than average (X _m ) are transform functions Is mapped to gray levels from X _m to X _L-1 . In the equation above, X 9 _m It can be seen that the re-mapping to X _m.

Therefore, when the histogram equalization of the low pass filtered input image is performed according to the calculated cumulative density function, the conversion function based on the cumulative density function is adjusted as shown in Equation 9 so that the average signal of the low pass filtered input image is mapped back to itself. A method of preventing the average brightness of a given image from occurring much during histogram equalization is called mean-matched histogram equalization in the present invention.

1 and 2, a preferred embodiment of the image quality improvement circuit using the mean-matching histogram equalization of the low-pass filtered signal according to the present invention will be described.

1 is a block diagram of an image quality improvement circuit using mean-matching histogram equalization of a low pass filtered signal according to the present invention.

In Fig. 1, the low pass filter 100 low pass filters the input signal Y (i, j) and outputs a low pass filtered signal A (i, j) = X _k .

The mean-matching histogram equalizer 200 improves the low-pass filtered signal A (i, j) output from the low-pass filter 100 by using an average-matching histogram equalization to improve the contrast of the low-pass filtered signal. Output the signal A _eq (i, j).

The subtractor 300 subtracts the low pass filtered signal A (i, j) output from the low pass filter 100 from the input signal Y (i, j).

The adder 400 adds the improved signal A _eq (i, j) output from the mean-matching histogram equalizer 200 and the output of the subtractor 300 to final output signal Y _H (i, j). ) This output signal Y _H (i, j) is represented by Equation 1 above.

FIG. 2 is a detailed circuit diagram of the circuit shown in FIG. 1 and will be described based on the average-matching histogram equalizer 200.

In FIG. 2, the frame histogram calculator 202 calculates the gray level distribution by inputting the low-pass filtered signal A (i, j) = X _k outputted from the low pass filter 100 in units of screens. Based on the gray level distribution, the probability density function p (X _k ) represented by Equation 6 is calculated. The screen unit may be a field, but here, it is a frame.

The CDF calculator 204 calculates the cumulative density function c (X _k ) using Equation 7 based on the probability density function p (X _k ) of one frame calculated by the frame histogram calculator 202. Calculate

The frame average calculator 206 calculates an average level X _m of the low-pass filtered signal of one frame. In this case, the frame average calculator 206 outputs the average level X _m calculated according to the synchronization signal SYNC to the CDF memories 208, the first and second mappers 210 and 212.

The CDF memory 208 updates the cumulative density function c (X _k ) calculated by the CDF calculator 204 in units of frames according to the synchronization signal SYNC, and performs a low pass among the cumulative density functions before one frame stored during the update. The cumulative density function value c (X _k ) corresponding to the filtered signal X _k and the mean cumulative density function value c (X _m ) with respect to the average level X _m output from the frame average calculator 206. )

The first mapper 210 may accumulate the cumulative density function value c (X _k ) corresponding to the low pass filtered signal X _k output from the CDF memory 208 and the cumulative density function value c (X (x) for the average level. _m )), the average level (X _m ) output from the frame average calculator 206 and the low pass filtered signal (X _k ) output from the low pass filter 100 are input to filter the low pass below the average level (X _m ). The signal X _k is mapped to a gray level from X ₀ to X _m using Equation 9 above.

The second mapper 212 has a cumulative density function value c (X _k ) corresponding to the low pass filtered signal X _k output from the CDF memory 208 and a cumulative density function value for the average level X _m . (c (X _m )), the average level (X _m ) output from the frame averaging calculator 206, and the low-pass filtered signal (X _k ) from the low pass filter (100) to input the equation (9). The low-pass filtered signal X _{k that} is above the average level X _m is mapped to a gray level from X _m to X _L-1 .

In this case, the low-pass filtered signal X _k input to the first and second mappers 210 and 212 is a signal of the next frame compared to the cumulative density function value output from the CDF memory 208.

Accordingly, the low-pass filtered signal output from the low pass filter 100 is input to input the low-pass filtered signal of the same frame as the cumulative density function value output from the CDF memory 208 to the first and second mappers 210 and 212. A frame memory delaying one frame may be configured. However, in the present invention, the hardware can be reduced by omitting the frame memory by using a characteristic of having a high correlation between adjacent frames.

The comparator 214 generates a selection control signal by comparing the low-pass filtered signal X _k and the average level X _m output from the frame average calculator 206.

The selector 216 selects the first mapper 210 when the low-pass filtered signal X _k is equal to or less than the average level X _m according to the selection control signal, and otherwise selects the second mapper 212. Here, frame histogram calculator 202 to selector 216 are referred to as mean-matched histogram equalizer 200.

In addition, without using the frame histogram calculator 202 and the CDF calculator 204 separately, a single gray scale distribution chart is calculated based on the gray level distribution for the low-pass filtered signal X _k . It may consist of blocks.

In addition, the present invention can be applied to a wide range of fields related to improving the image quality of a video signal. That is, the present invention can be applied to broadcasting devices, radar signal processing systems, medical engineering, home appliances, and the like.

As described above, the present invention provides a low-pass filtering of the input signal and adjusts the conversion function such that the average gray level of the low-pass filtered signal is mapped to itself when the histogram equalizes according to the cumulative density function of the low-pass filtered signal. At the same time, the image quality can be improved by suppressing the improvement of the noise when the contrast is improved.

Claims (11)

A method for improving image quality of a video signal represented by a predetermined number of gray levels by using histogram equalization: (a) low pass filtering the input video signal to output a low pass filtered signal; (b) outputting an improved signal by performing a histogram equalization using the cumulative density function of the low pass filtered signal as a conversion function and adjusting the conversion function such that the average level of the low pass filtered signal is mapped to itself; obtaining a difference from the low pass filtered signal from the input image signal; And and (d) adding the obtained difference value to the improved signal to output an output signal. The method of claim 1, wherein step (b) obtaining a cumulative density function by inputting the low-pass filtered signal on a screen basis; (b2) calculating an average level by inputting the low-pass filtered signal in units of screens; And (b3) converting the level of the low-pass filtered signal using the obtained cumulative density function as a conversion function, and outputting an improved signal by adjusting the conversion function such that the obtained average level is mapped to the average level as it is. Image quality improvement method characterized in that. The image quality improving method according to claim 1, further comprising the step (b31) of delaying the low-pass filtered signal in units of screens and outputting the delayed signal in the step (b3). 3. The method of claim 2, wherein the difference between the output signal and the enhanced signal is equal to the difference between the input video signal and the low pass filtered signal. A circuit for improving image quality of a video signal represented by a predetermined number of gray levels by using histogram equalization: Low pass filtering means for low pass filtering the input video signal and outputting a low pass filtered signal; A mean-matching histogram equalization means for outputting an improved signal by adjusting the transform function such that the cumulative density function of the low-pass filtered signal is used as a transform function and the average level of the low-pass filtered signal is mapped to itself. ; Detection means for detecting a difference between the low-pass filtered signal in the input video signal; And And adding means for adding the detected difference value and the improved signal to output an output signal. 6. The apparatus of claim 5, wherein the mean-matching histogram equalization means First calculating means for inputting the low-pass filtered signal in units of screens to calculate an average level; Second calculation means for calculating the cumulative density function value of the screen unit and the cumulative density function value for the average level based on the gray level distribution chart calculated by inputting the low-pass filtered signal in the screen unit; ; And The gray level of the low-pass filtered signal is converted according to the cumulative density function value corresponding to the low-pass filtered signal and the cumulative density function value of the average level, and the average level is mapped to the average level as it is, thereby improving the signal. And an output means for outputting. The display apparatus of claim 6, further comprising a screen memory for delaying the low-pass filtered signal by one screen unit to input a signal having the same frame and the same density calculated by the second calculation means to the output means. Characteristic improvement circuit characterized in that. The image quality improvement circuit according to claim 7, wherein the screen unit is a frame unit. The method according to claim 6, wherein the cumulative density function value calculated by the second calculation means is updated in units of screens, and the cumulative density function value stored during the update and the cumulative density function value for the average level are output to the output means. The image quality improvement circuit further comprises a buffer. The method of claim 6, wherein the output means Using the obtained cumulative density function as a conversion function, the conversion function is adjusted to map the average level calculated by the first calculation means to the average level as it is, and the low-pass filtered signal below the average level has a corresponding cumulative density function value. And a first mapper that maps to a gray level of a first range according to a cumulative density function value for the average level. A second mapper for mapping the low-pass filtered signal larger than the average level calculated by the first calculation means to a gray level of a second range according to a corresponding cumulative density function value and a cumulative density function value for the average level; A comparator for comparing the low-pass filtered signal with the average level to generate a selection control signal; And And a selector for selecting a first mapper if the low-pass filtered signal is less than the average level according to the selection control signal, and otherwise selecting a second mapper. The method of claim 7, wherein the output means By using the obtained cumulative density function as a conversion function, the conversion function is adjusted so that the average level calculated by the first calculation means is mapped to the average level as it is, and the low pass filtered signal which is less than or equal to the average level output from the screen memory corresponds thereto. A first mapper for mapping to a gray level of a first range according to a cumulative density function value and a cumulative density function value for the average level; A second mapper for mapping the low-pass filtered signal larger than the average level output from the screen memory to a gray level in a second range according to a corresponding cumulative density function value and a cumulative density function value for the average level; A comparator for generating a selection control signal by comparing the low pass filtered signal output from the screen memory with the average level; And And a selector for selecting a first mapper if the low-pass filtered signal output from the screen memory is less than the average level according to the selection control signal, and otherwise selecting a second mapper.